Highly reliable distributed file systems employ multiplexing, which is to create a replication file system having the same content as a master file system and store the replication file system into a computer node other than a computer node storing the master file system. These distributed file systems update the master file system and reflect the updated content on the replication file system to maintain file system multiplexing. Even if either the master file system or the replication file system crashes, these distributed file systems can restore the content of the file system from the other file system. Moreover, by creating a plurality of replication file systems, it is possible to secure higher reliability.
An example of a method for creating a replica of master data into a plurality of computer nodes is shown in FIG. 10 of Patent Document 1. In the technique described in Patent Document 1, the computer nodes having the replication data are connected in a row. When the data on the master computer node is updated, the replication data on one computer node at the head of the computer nodes having the replication data is also updated. Next, the updated data is transferred from the head computer node to a second computer node, and the replication data on the second computer node is updated. The same operation is repeatedly performed in order up to the last computer node. Such multiplexing in which data necessary for updating replication data is relayed from a computer node to a computer node is called relay multiplexing. Moreover, the side of the master computer side is called an upstream side, and the side of the last computer node is a downstream side.
[Patent Document 1] Japanese Unexamined Patent Application Publication No. JP-A 2007-200086
The relay multiplexing described above makes it possible to prevent concentration of load on a specific computer node. However, when a delay occurs in communication between the computer nodes, a large amount of data that cannot be sent out to the downstream side remains in a storage part of the computer node on the upstream side from a place in which the delay occurs.